AGENT SKILL

plugin-test

Run NCC plugin E2E tests. Supports quick (4 projects), full (4+monorepo+modifications), install-test (npx vs marketplace with non-spec-dev focus), or custom test modes. Validates spec→dev flow, skills, hooks, code quality, security, a11y, and token efficiency.

使用场景测试与质量
License 许可
MIT
最近更新
2026-03-27
Stars
★ 3
Forks
1

01 前言 Front Matter

name
plugin-test
agent
plugin-test-orchestrator
context
fork
description
Run NCC plugin E2E tests. Supports quick (4 projects), full (4+monorepo+modifications), install-test (npx vs marketplace with non-spec-dev focus), or custom test modes. Validates spec→dev flow, skills, hooks, code quality, security, a11y, and token efficiency.
argument-hint
--quick | --full | --install-test | --phase1 | --phase2 | --security | --a11y | --codequality | [project-name,...] | (no args = suggest mode)

02 资源文件 7

03 SKILL.md 正文

/plugin-test — NCC Plugin E2E Test Suite

Usage

/plugin-test                    # Interactive — suggest a test mode based on recent changes
/plugin-test --quick            # Phase 1 only: 4 diverse projects × 4 features each
/plugin-test --full             # Phase 1 + Phase 2: full suite including monorepo + cross-feature mods
/plugin-test --install-test     # Install test: 2 projects (npx + marketplace), 5 features each, non-spec-dev focus
/plugin-test --phase1           # Phase 1 only (same as --quick)
/plugin-test --phase2           # Phase 2 only: 10-feature monorepo (solo + team)
/plugin-test --security         # Security-focused: run security-reviewer on all generated code
/plugin-test --a11y             # Accessibility-focused: run browser-tester a11y checks on UI projects
/plugin-test --codequality      # Code quality audit: deep analysis of all generated source files
/plugin-test --token             # Token optimization audit: run token-optimizer on plugin source
/plugin-test healthcare,marketplace   # Run specific projects only
/plugin-test --create              # /create pipeline test: new + existing project ideation

Test Modes

Phase 1: Quick Validation (--quick, --phase1)

4 diverse projects, each with 4 features + extra agent verification.

Project Stack Extra Agents
healthcare Prisma + next-auth + zod + vitest /security, /review
design-system framer-motion + playwright + storybook browser-tester, /qa --a11y
realtime-collab drizzle + tanstack-query + zustand + msw performance-optimizer, /loop
marketplace Prisma + stripe + ai-sdk + @vercel/blob /security --diff, /review, /status

Per-project: NCC install → /init → 4× spec→dev (3 tasks each) → extra agent runs → TEST_LOG.md

Reviewer scores 15 categories: 1-10: Installation, Hooks, Skills, Spec, Design, PLAN, TDD, Token, Code, Flow 11: Rules size (<200 lines each) 12: Skill selection accuracy (manifest vs package.json) 13: Extra agent execution results 14: Document sync (CONTEXT.md/design.md consistency) 15: Code quality detail (error handling, Zod, composition, Server/Client)

Install Test (--install-test)

Tests the published version of NCC using two real installation methods. Primary focus: non-spec-dev features.

Project Install Method Stack Features Non-Spec-Dev Focus
npx-healthcare npx nextjs-claude-code@latest Prisma + next-auth + zod + vitest 5 features /init, /security, /log, /commit, /rule, /test, hooks, skills
mkt-marketplace /plugin marketplace add + /plugin install Prisma + stripe + ai-sdk + @vercel/blob 5 features /init, /review, /status, /debug, /qa, /loop, /commit, hooks, skills

Per-project: Install → /init → 5× spec→dev (2-3 tasks each) → non-spec-dev tests → TEST_LOG.md

Reviewer scores 20 categories (base 15 + 5 install-test categories): 16: Init Completeness (PROJECT.md, ARCHITECTURE.md, STATE.md quality) 17: Security & Log Setup (SECURITY_STRATEGY.md, LOG_STRATEGY.md, audit results) 18: Git Workflow (/commit conventional format + REQ linking) 19: Debug & Loop (DEBUG.md hypothesis-driven, LOOP_NOTES.md tracking, /status read-only) 20: Installation Method Fidelity (npx: local files + manifest; marketplace: namespace + CLAUDE_PLUGIN_ROOT)

Phase 2: Large-scale Monorepo (--phase2)

2 identical CRM projects (solo vs team mode), each with:

  • 10 features in dependency chain
  • 5 cross-feature modifications (api-layer error format, RBAC expansion, tagging, cursor pagination, custom stages)
  • Document Sync verification (CONTEXT.md, design.md, history entries)
  • Solo vs Team comparison

Features: api-layer → auth-system → contact-management → deal-pipeline → email-integration → activity-feed → reporting → team-workspace → notification-center → admin-settings

/create Pipeline Test (--create)

Tests the /create ideation pipeline on both new and existing projects.

Project Scenario Validation
create-new Empty project → /init → /create "AI recipe app" VISION.md, C-REVIEW.md, DECISION.md exist in spec/create/
create-existing Healthcare project → /create "notification system" Reads spec/PROJECT.md context, outputs to spec/create/

Per-project: NCC install → /init → /create → verify output docs → token isolation check → spec conversion test

Reviewer checks 5 categories:

  1. Document Quality: VISION.md has Problem/Demand/Target/Value sections
  2. C-Level Coverage: C-REVIEW.md has all 5 assessments (CEO/CTO/CPO/CMO/CDO) with verdicts
  3. Decision Clarity: DECISION.md has chosen approach + key decisions
  4. Token Isolation: After /create, run /spec and verify spec-writer does NOT read spec/create/
  5. Spec Conversion: If converted, output is valid /spec command format

Full Suite (--full)

Runs Phase 1, analyzes issues, then runs Phase 2. Generates consolidated REPORT with cross-phase comparison.

Security Audit (--security)

For each generated project:

  • Read installed security-reviewer agent
  • Scan all API routes for OWASP Top 10: injection, auth bypass, sensitive data exposure, CSRF
  • Check MSW handlers for auth mock bypass patterns
  • Check Zod validation coverage
  • Output: SECURITY_REVIEW.md per project

Accessibility Audit (--a11y)

For UI-heavy projects (design-system or any with UI components):

  • Read installed browser-tester agent
  • WCAG 2.1 AA checklist per component
  • Check: ARIA attributes, keyboard nav, focus management, color contrast
  • Check: prefers-reduced-motion respect, semantic HTML, role attributes
  • Output: ACCESSIBILITY_REVIEW.md per project

Code Quality Audit (--codequality)

Deep source code analysis across all projects:

  • Error handling grade (try/catch, error classification, no silent swallowing)
  • Zod validation grade (boundary validation, schema completeness)
  • TypeScript strictness (no any, no as unknown, proper inference)
  • Composition patterns (Context vs prop drilling, compound components)
  • Server/Client discipline ('use client' minimization)
  • DRY compliance (no duplicate logic across routes)
  • Named constants (no magic numbers/strings)
  • Output: code quality table in REPORT

Review Checklist

Standard (15 categories, /15 scoring)

# Category What to check
1 Installation agents 25+, scripts 6+, skills 5+, rules 3, spec/rules 8+, CLAUDE.md
2 Settings & Hooks env keys, SessionStart(2), PreToolUse(2), PostToolUse(3), Stop(1)
3 Skill Selection core skills vs package.json library matching
4 Spec Quality frontmatter (mock, testing, deps), REQ-NNN:, all required sections
5 Design Quality Components, State, Data Flow, Technical Decisions
6 PLAN.md Quality domain tags, model, wave, MSW tasks, Approved-at, Budget, Team Composition (team mode)
7 TDD/Mock MSW handlers for all endpoints, mock:true enforcement, test files
8 Token Efficiency file sizes <200 lines, artifact limits, skill budget rules
9 Code Quality try/catch, Zod, strict TS, no stubs, Server/Client
10 Flow Completeness STATE.md, CONTEXT.md, [x] tasks, TEST_LOG.md
11 Rules Size all spec/rules/ <200 lines, no redundancy
12 Skill Accuracy manifest vs detected libs, on-demand coverage
13 Extra Agents security/a11y/performance/review/loop results
14 Document Sync CONTEXT.md consistent after modifications, no stale decisions
15 Code Quality Detail error handling grade, composition, naming, DRY

Install Test Extended (20 categories, /20 scoring)

All 15 standard categories plus:

# Category What to check
16 Init Completeness PROJECT.md framework detection, ARCHITECTURE.md structure, STATE.md features
17 Security & Log Setup SECURITY_STRATEGY.md, LOG_STRATEGY.md, audit findings quality
18 Git Workflow /commit conventional format, REQ linking, no git add -A
19 Debug & Loop DEBUG.md hypotheses, LOOP_NOTES.md tracking, /status read-only
20 Install Method Fidelity npx: local files + manifest + doctor; marketplace: namespace + plugin paths

Report Output

Test artifacts are written to __plugin-tests__/ during execution:

  • Per-project: TEST_LOG.md, REVIEW.md
  • Extra agents: SECURITY_REVIEW.md, ACCESSIBILITY_REVIEW.md, PERFORMANCE_REVIEW.md, REVIEW_RESULT.md

After each test run, generate a key-summary English report and save it to reports/REPORT_YYYYMMDD_HHMM.md (inside this skill directory). The report should contain:

  • Date, branch, version, scope
  • Score summary table
  • Key findings (bullet points, not full prose)
  • Critical issues with severity
  • Recommendations
  • For --install-test: installation method comparison table + non-spec-dev feature results

Keep each report under 80 lines. This is the archival summary — the raw artifacts in __plugin-tests__/ have full details.

Past Reports

Reports are stored in reports/ directory. By default, read only the latest report (highest datetime in filename) to understand current plugin quality baseline. Read older reports only when the user explicitly asks for trend analysis or historical comparison.

Report Scope Score
REPORT_20260321_1900.md Main branch validation (2 projects) 9.5/10 avg
REPORT_20260321_1700.md R5: Incremental fix-and-retest (3 projects) 8.7/10 avg
REPORT_20260321_1500.md R4: Full suite Phase 1+2 (4+2 projects, 15 categories) 12.0/15 avg
REPORT_20260321_1300.md R3: Monorepo solo vs team (10 features × 2 modes) Solo 6/10, Team 5/10
REPORT_20260321_1100.md R2: 4 projects × 4 features (post-fix) 9.0/10 avg
REPORT_20260321_0900.md R1: 4 projects × 1 feature (baseline) 7.5/10 avg

NCC Source Path

For dev/Phase 1/Phase 2 modes: node {NCC_SOURCE}/dist/index.js (tests current branch build). For install-test mode: uses published npm version (npx nextjs-claude-code@latest) and marketplace.

04 同仓库更多 Skills

ByeongminLee
ByeongminLee/nextjs-claude-code

zustand

Zustand state management guide. Use when working with store code (src/store/**), implementing actions, managing state, or creating slices. Triggers on Zustand store development, state management questions, or action implementation.

★ 3
ByeongminLee
ByeongminLee/nextjs-claude-code

web-vitals

暂无简介

★ 3
ByeongminLee
ByeongminLee/nextjs-claude-code

vercel-storage

Vercel storage expert guidance — Blob, Edge Config, and Marketplace storage (Neon Postgres, Upstash Redis). Use when choosing, configuring, or using data storage with Vercel applications.

后端与接口
★ 3
ByeongminLee
ByeongminLee/nextjs-claude-code

vercel-functions

Vercel Functions expert guidance — Serverless Functions, Edge Functions, Fluid Compute, streaming, Cron Jobs, and runtime configuration. Use when configuring, debugging, or optimizing server-side code running on Vercel.

开发工程师后端与接口
★ 3
ByeongminLee
ByeongminLee/nextjs-claude-code

ui-ux-pro-max

UI/UX design intelligence for web and mobile. Includes 50+ styles, 161 color palettes, 57 font pairings, 161 product types, 99 UX guidelines, and 25 chart types across 10 stacks (React, Next.js, Vue, Svelte, SwiftUI, React Native, Flutter, Tailwind, shadcn/ui, and HTML/CSS). Actions: plan, build, create, design, implement, review, fix, improve, optimize, enhance, refactor, and check UI/UX code. Projects: website, landing page, dashboard, admin panel, e-commerce, SaaS, portfolio, blog, and mobile app. Elements: button, modal, navbar, sidebar, card, table, form, and chart. Styles: glassmorphism, claymorphism, minimalism, brutalism, neumorphism, bento grid, dark mode, responsive, skeuomorphism, and flat design. Topics: color systems, accessibility, animation, layout, typography, font pairing, spacing, interaction states, shadow, and gradient. Integrations: shadcn/ui MCP for component search and examples.

设计师数据分析师前端开发视觉设计
★ 3
ByeongminLee
ByeongminLee/nextjs-claude-code

turborepo

Turborepo monorepo build system guidance. Triggers on: turbo.json, task pipelines, dependsOn, caching, remote cache, the "turbo" CLI, --filter, --affected, CI optimization, environment variables, internal packages, monorepo structure/best practices, and boundaries. Use when user: configures tasks/workflows/pipelines, creates packages, sets up monorepo, shares code between apps, runs changed/affected packages, debugs cache, or has apps/packages directories.

★ 3